GNSS dynamic positioning model displacement detection method based on automatic encoder

Abstract

The invention belongs to the technical field of deformation monitoring based on GNSS dynamic positioning, and discloses a GNSS dynamic positioning model displacement detection method based on an automatic encoder. Firstly, de-noising processing and time domain difference are carried out on a coordinate time sequence; secondly, normalizing and serializing the differentiated coordinate time sequence; then, constructing an automatic encoder abnormal value detection model and training the automatic encoder abnormal value detection model; thirdly, a reconstruction value is obtained through the trained model, the mean absolute error (MAE) is calculated, and the maximum MAE value serves as a reconstruction loss threshold value delta MAX to be used for detecting whether displacement occurs or not; and finally, an MAE loss value delta is obtained by using the trained model, and if the delta exceeds the calculated threshold value delta MAX, the delta is marked as an abnormal value, that is, the current epoch is displaced. According to the method, the problem that GNSS displacement detection with the characteristics of large noise and small displacement is difficult is solved, displacement can be effectively detected from a dynamic positioning result, and the detection displacement threshold value is dynamically calculated through data without manual intervention.

Images